Comparisons

Academi.cx vs Walter Writes AI Accuracy Test

We compared each detector’s overall AI scores and sentence-level highlights with Turnitin’s results across six essays.

Brayden · 2 min read

Walter Writes AI pairs an AI humanizer with a detector marketed as a fast, reliable way to check writing before submission. It promises high accuracy while minimizing false positives, positioning itself as a tool students and educators can trust. We compare its detector with Academi.cx to find out which gives a closer indication of what Turnitin will report.

Watch the full comparison on YouTube

How the test worked

Six essays were sourced from IvyPanda, a free essay database with no affiliation to Academi.cx. Selected paragraphs in each essay were rewritten using ChatGPT to introduce AI-generated content. The same version of each document was then submitted to Academi.cx, Walter Writes AI, and Turnitin. We compared how closely Academi.cx and Walter Writes AI matched Turnitin’s overall AI scores and sentence-level highlights. Two of the essays landed in Turnitin's under-20% zone, where it only shows an asterisk instead of a real number, so an extra document was added to get another clean data point with visible highlights.

The results from each detector

Six-document score comparison. Document 1: Academi.cx 67%, Walter Writes 15%, Turnitin 54%. Document 2: 0%, 5%, and 0%. Document 3: 26%, 28%, and hidden below 20%. Document 4: 28%, 25%, and 29%. Document 5: 7%, 30%, and hidden below 20%. Document 6: 30%, 12%, and 33%. Capped blue lines show Academi.cx predicted Turnitin ranges.
  • Document 1: Academi.cx detected 67%, predicting 60 to 71%. Turnitin returned 54%, 6 points below the range. Walter Writes predicted 15%, 39 points below Turnitin.
  • Document 2: Academi.cx detected 0%, predicting 0 to 19%. Turnitin also returned 0%. Walter Writes predicted 5%, and the round was a no contest.
  • Document 3: Academi.cx detected 26%, predicting 18 to 30%. Turnitin's exact score was hidden below 20%. Walter Writes predicted 28%.
  • Document 4: Academi.cx detected 28%, predicting 22 to 32%. Turnitin returned 29%, inside the range. Walter Writes predicted 25%, 4 points below Turnitin.
  • Document 5: Academi.cx detected 7%, predicting 0 to 19%. Turnitin's exact score was hidden below 20%. Walter Writes predicted 30%.
  • Document 6: Academi.cx detected 30%, predicting 24 to 34%. Turnitin returned 33%, inside the range. Walter Writes predicted 12%, 21 points below Turnitin.

Two very different detectors

We also looked at how much detail each detector gave students about their results. Walter Writes reports results as a percentage chance of being human generated rather than an AI score, and on the very first document, it didn't show any sentence highlights at all, a pattern that showed up again more than once. Academi.cx, by comparison, displayed a graded likelihood on every sentence and paragraph, in every document, giving a much clearer picture of what would actually get flagged. A missing percentage or a percentage with nothing behind it doesn't tell a student much about what to actually go back and change.

Average deviation from Turnitin: Academi.cx, shown in blue, 4.25 percentage points; Walter Writes AI, shown in dark purple, 17.25 percentage points. Values are mean absolute differences based on Documents 1, 2, 4, and 6. Documents 3 and 5 are excluded because their exact Turnitin scores are hidden.

The final scorecard

Academi.cx took the comparison 5 to 0, with one round called a no contest due to almost no AI content on either side. Walter Writes markets itself as the top AI detector available, but across six documents its predictions swung as much as 39 points from Turnitin's actual score, and its sentence highlights were inconsistent at best. Academi.cx's predicted ranges captured Turnitin's actual score on three of the four documents where an exact score was available, and its highlights lined up closely with what Turnitin flagged throughout.

Related reading